SemiAnalysis has launched GDPx, a new benchmark designed to measure how well frontier AI models perform as autonomous agents across the most economically valuable domains. The research firm, known for its semiconductor and AI industry analysis, published the benchmark on its newsletter, with contributions from Max Kan, Andrew Megalaa and Dylan Patel.
The headline result is humbling for the industry: Opus 5.5, the current leader, scores just 12%. GDPx spans 11 domains including chip design, cybersecurity, recursive self-improvement (RSI), hardware design, biology, finance, legal work, quantitative trading, software engineering, materials science and alignment. SemiAnalysis says it wants GDPx to become the world's best benchmark for measuring general agentic capabilities.
What GDPx measures: 11 domains
Unlike benchmarks built around exam-style questions, GDPx is an agentic benchmark. That means it tests whether a model can complete multi-step, real-world tasks in fields where companies and governments actually spend money today.
SemiAnalysis argues the first trillion dollars of AI revenue will concentrate in sectors that can deploy frontier intelligence at high return on investment, rather than spreading evenly across the economy. GDPx is built to track progress in exactly those sectors.
- RSI (recursive self-improvement)
- Chip design
- Hardware design
- Cybersecurity
